Scale the TiFlash Cluster

This document describes how to scale in and out nodes in the TiFlash cluster.

Scale out a TiFlash node

The following example shows how to scale out a TiFlash node if you deploy TiFlash on the 192.168.1.1 node.

Edit the inventory.ini file to add the TiFlash node information:
Currently, you can only add IP but not domain name.
```
[tiflash_servers]
192.168.1.1
```

Edit the hosts.ini file to add the node information:

[servers]
192.168.1.1
[all:vars]
username = tidb
ntp_server = pool.ntp.org

Initialize the new node:
- Configure the SSH mutual trust and sudo rules on the Control Machine:
```
ansible-playbook -i hosts.ini create_users.yml -l 192.168.1.1 -u root -k
```
- Install the NTP service on the target machine:
```
ansible-playbook -i hosts.ini deploy_ntp.yml -u tidb -b
```
- Initialize the node on the target machine:
```
ansible-playbook bootstrap.yml -l 192.168.1.1
```

Deploy the new node:

ansible-playbook deploy.yml -l 192.168.1.1

Start the new node:

ansible-playbook start.yml -l 192.168.1.1

Update the configuration of Prometheus and restart it:

ansible-playbook rolling_update_monitor.yml --tags=prometheus

Access the monitoring platform using your browser, and view the status of the cluster and the new node.

Scale in a TiFlash node

The following example shows how to scale in a TiFlash node if you stop the TiFlash service on the 192.168.1.1 node.

Note

The offline process described in this section does not delete the data on the offline node. If you need to take the node online again, delete the data manually.

Take down the node to be scaled in. See Take a TiFlash node down for details.
Check whether the node has been offline successfully using Grafana or pd-ctl (the offline process takes some time).
After the store corresponding to TiFlash disappears, or the state_name becomes Tombstone, execute the following command to shutdown the TiFlash process:
```
ansible-playbook stop.yml -l 192.168.1.1
```
If the node still has other services and you want to stop TiFlash only, use the following command:
```
ansible-playbook stop.yml -t tiflash -l 192.168.1.1
```
Edit the inventory.ini and hosts.ini files to remove the node information.

Update the configuration of Prometheus and restart it:

ansible-playbook rolling_update_monitor.yml --tags=prometheus

Access the monitoring platform using your browser, and view the status of the cluster.